• We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Status
Not open for further replies.

CynicalRyan

Captain
30 Badges
May 26, 2009
447
0
  • Hearts of Iron II: Armageddon
  • Europa Universalis III Complete
  • Divine Wind
  • Hearts of Iron III
  • Heir to the Throne
  • Semper Fi
  • Europa Universalis: Rome
  • Cities: Skylines - Green Cities
  • Mount & Blade: Warband
  • Cities: Skylines - After Dark
  • Cities: Skylines - Snowfall
  • Cities: Skylines - Natural Disasters
  • Cities: Skylines - Mass Transit
  • Surviving Mars
  • Victoria 2
  • Cities: Skylines - Parklife Pre-Order
  • Cities: Skylines - Parklife
  • Cities: Skylines Industries
  • Cities: Skylines - Campus
  • Cities: Skylines
  • 500k Club
  • Rome: Vae Victis
  • Rome Gold
  • Victoria: Revolutions
  • Europa Universalis III Complete
  • Europa Universalis III Complete
  • Europa Universalis III: Chronicles
  • Europa Universalis III
  • Crusader Kings II
  • Cities in Motion
Will Hearts of Iron 3 use Goal-Oriented Action Planning (or a variant thereof)?

GOAP website with in-depth articles.

In a nutshell: Instead of telling the AI word-for-word how it should go about solving a task, it gets an arsenal of actions allowing it to solve the task in different ways, depending on the circumstances.

AFAIK, Empire: Total War uses this concept (not just F.E.A.R. :p) to great success (haven't played it, so I wouldn't know how well it works in strategy games, but the AI blows me away in FEAR2, sometimes quite literally).

Second question: If you didn't know about this (and/or don't want to delay HoI3), will HoI4 (or EU4, or...) use this system? :p
 
As far as I can tell it is a buzzword. Of course the Paradox AI functions exactly in the way you described (as all you ever tell it is target provinces and overall stance), but the architecture will naturally be different from GOAP.
 
As far as I can tell it is a buzzword.
Far from it, as my first-hand experience with F.E.A.R. 2 shows me, as well as the research of the "predecessor" in academia.

Of course the Paradox AI functions exactly in the way you described (as all you ever tell it is target provinces and overall stance), but the architecture will naturally be different from GOAP.
Well, looking at the AI files from current Paradox games, it's pretty much a state-machine, and thus predictable for an experienced player.
 
AFAIK, Empire: Total War uses this concept (not just F.E.A.R. :p) to great success (haven't played it, so I wouldn't know how well it works in strategy games,

Sorry, I don't have time to talk about our AI atm, but I was not aware of there being an AI in Empire Total War.

But yeah, we've been using goal-oriented plans driving our ai's since eu1. You can't determine hiw the AI in the older engine works by the cripple-scripts that it has to make the AI just do historical things.
 
  • 1
  • 1
Reactions:
Sorry, I don't have time to talk about our AI atm, but I was not aware of there being an AI in Empire Total War.
Well, it might be something for the Developer Diaries (esp. if we, as players, can make use of it thanks to automation). Thank you, anyway. :)
 
If you play enough multi player, you do guess many things what player does, as they are best choices.
Well, yes. But that's a far cry from "GER won't invade Russia, if GER has less than 75% of Russia's strenght".

IOW, it's not guessing what the best / most likely strategy is, but rather knowing how the AI will react if I present it with a specific input.
 
AFAIK, Empire: Total War uses this concept (not just F.E.A.R. :p) to great success

Lol! Yeah right ... :confused: HOI3 could ship with no AI and it would probably be smarter then E:TW.

Not much has been mentioned about AI except 'its much better' so I'm eagly waiting the mechanics of the AI :p Particuarlly to see how well they adapt to changing conditions, something that was poorly handled in HOI2.
 
Far from it, as my first-hand experience with F.E.A.R. 2 shows me, as well as the research of the "predecessor" in academia.

What I meant to say is that it's not really some particular, well-defined thing. I assume that the AI opponents in Empire: Total War behave differently (under the hood) from the Enclave propaganda droids I used to to practice my marksmanship on in Fallout 3 :D

Well, looking at the AI files from current Paradox games, it's pretty much a state-machine, and thus predictable for an experienced player.

The EU3/Rome AIs were almost entirely unscripted, so we know very little about the actual contents.

However, how exactly is GOAP different from a state machine? Just looking at the definitions on Wiki, I fail to grasp how an AI could not be a state-machine...
 
Well, yes. But that's a far cry from "GER won't invade Russia, if GER has less than 75% of Russia's strenght".

IOW, it's not guessing what the best / most likely strategy is, but rather knowing how the AI will react if I present it with a specific input.

Well, provided than AI solves an optimization problem every time you present it with specific conditions, it will always be predictable to a large degree even if you don't know the function it's optimizing. I don't see a way around it, and as pointed out, you can do the same thing with your human opponents, provided that they actually try to plan their actions systematically instead of making random moves.
 
On a sidenote:
Im curious.
I may post in this thread, but in cant post in Johans poll about the next video.
I also can not vote.
Why?

I think the forum is setup that in polls only registered customers can vote.

and I THINK that this forum has copied the settings from eu3 for now.

will be fixed i think.
 
But yeah, we've been using goal-oriented plans driving our ai's since eu1. You can't determine hiw the AI in the older engine works by the cripple-scripts that it has to make the AI just do historical things.

So...I assume that the scripting is reduced in HoI3 compared to HoI2? (Please tell us it is...)

I assume that things like the inherent dislike of certain countries for certain alliances (which could not be changed even by bringing the relations to 200) is now directly modeled through the "triangle," for example?
 
What I meant to say is that it's not really some particular, well-defined thing. I assume that the AI opponents in Empire: Total War behave differently (under the hood) from the Enclave propaganda droids I used to to practice my marksmanship on in Fallout 3 :D
Actually: No, Fallout 3 uses GOAP, too. :p

However, how exactly is GOAP different from a state machine? Just looking at the definitions on Wiki, I fail to grasp how an AI could not be a state-machine...

The main difference is as such:
A state-machine AI goes through the motions every time, and needs entry and exit points for a task. Also, everything the AI is supposed to be capable of needs to be implemented before hand, and it has to be "taught" new skills, and the state-machine(s) have ot be updated if you include a new action (for example, if you have dark rooms, the AI reacts in a specific state. If it is supposed to switch on the light, the AI has to be updated to do that).

GOAP uses goals and task to accomplish the same. Explanation by way of example:

Goal: Kill enemy
Task: Aquire weapon
Task: Find enemy
Task: Fire weapon

Goal: Ensure own survival
Task: Find cover
Task: Use cover
Task: heal thyself

With GOAP, the AI can fullfil these two goals simultaneous, without having to exit and enter a given state.

In short: The AI will try to kill you, while it is seeking cover. In a state-machine, it does one or the other, or becomes a maintenance nightmare for developers.

This can lead to all sorts of interesting emerging behavior. FEAR's AI, for example, uses the above goals in such a way, that the AI seems to flank the player and/or employs pincer movements.

I honestly can't explain it as well as the papers linked in my original post, so I gotta refer you to them.
 
Status
Not open for further replies.